AI customer feedback analysis for better product decisions
Support conversations and public reviews become one weekly report grouped by theme, so product teams can see what customers keep saying rather than react to the latest complaint.
The situation
Feedback arrived in two places that never met. Support tickets held problems, questions, and complaints from paying customers. Public reviews on Shopify and Trustpilot showed a different sample: customers choosing to share their experience publicly.
Both sources were useful, but neither was easy to read as a whole. One support conversation can contain an important issue. Six reviews describing the same issue tell a different story. Nobody was going to find either pattern by reading everything every week.
The company knew those messages and reviews contained useful evidence for product-roadmap decisions, feature improvements, and bug fixes. It needed a reliable way to turn that evidence into a shared picture.
The constraint
The obvious build is a real-time feed: classify each ticket and review as it arrives, then push it somewhere. That produces a stream nobody reads and encourages teams to react to the most recent comment rather than the issue that keeps appearing.
The weekly cadence was deliberate, not a technical convenience. A week is long enough that one loud complaint does not distort the picture and short enough that a growing issue remains visible while it is still cheap to fix.
Collecting public reviews was its own challenge. Review sources must be collected reliably, checked against what is already stored, and monitored for failures. A weekly report that silently loses one source is worse than one that fails visibly.
The report needed a fixed structure too. A model asked to write a weekly summary will change the headings, order, and level of detail from week to week. That makes comparisons difficult. The output therefore uses a fixed Notion template; the model fills the structure rather than inventing it.
What we built
Support tickets are categorised as they are created. A scheduled process collects new public reviews each week and checks them against stored records so nothing is counted twice. Both sources feed the same weekly analysis, which groups the previous week’s feedback into themes rather than listing every message.
The report is written into Notion using the same template every week. It shows recurring problems, so the team can tell whether an issue affects one account or several, and recurring praise, so product teams can see which parts of the experience to protect.
We also built a visual dashboard so the product team could see which topics were mentioned most each week.
Process diagram
Where it stands
Product and customer-facing teams work from the same weekly picture rather than separate anecdotes. An issue affecting several customers appears as a theme instead of three unconnected tickets nobody joined together.
Next
Lead generation for recruitment agencies
Ready to boost business results with AI automation?
Book a 30-minute operations review. We will look at where manual work is limiting your operation and whether there is a practical opportunity to create capacity.
Book a 30-minute operations reviewSwiss precision. Valley speed.